Statistical evidence of accidents prevention and costs reduction, through alcohol and drug testing at work. - an observational study

Size: px

Start display at page:

Download "Statistical evidence of accidents prevention and costs reduction, through alcohol and drug testing at work. - an observational study"

Marilynn Floyd
8 years ago
Views:

1 Statistical evidence of accidents prevention and costs reduction, through alcohol and drug testing at work - an observational study

2 Paulo Henriques dos Marques, PhD

3 Prior knowledge derived from literature review Alcohol and drugs abuse causes risks of accidents

4 Prior knowledge derived from literature review Thus, in order to prevent accidents, tests for alcohol and drugs are performed

5 Prior knowledge derived from literature review Programmes for testing alcohol and drugs (A&D) at the workplace, at random and by surprise, are believed to have a positive impact on safety and to reduce individual s accident risk Despite this perception, there is limited scientific evidence and poor statistical support of this assumption Another issue which has not been properly answered yet - in the rare publications reporting frequency of testing, it still remains to be confirmed the existence of a specific frequency that could be more preventive

scientific evidence and poor statistical support of this assumption Another issue which has not been properly answered yet - in the rare

6 Derived research hypotheses This study aimed at testing whether there is such a causeeffect relationship between A&D testing and post-accident reduction, and how to quantify it

7 To fill in these gaps, this study raised two hypotheses for Hipótese de variação dos acidentados em função da sujeição a testes sem acidentes prévios research: 12 H1 (preventive effect): The frequency of alcohol and drug testing is negatively associated with the incidence rate of accidents occurred after the tests H2 (optimal frequency): Frequência relativa de acidentados após testes ]0,0; 0,5] ]0,5; 1,0] ]1,0; 2,0] ]2,0; 4,0] ]4,0; 8,0] > 8,0 > 8,0 > 8,0 Intensidade de sujeição a testes sem acidentes prévios There is an optimal frequency of tests and postaccidents that represents the most efficient frequency, beyond which, increasing the number of annual tests will result in marginal variation of accidents

de acidentados após testes 10 8 6 4 2 0 0 ]0,0; 0,5] ]0,5; 1,0] ]1,0; 2,0] ]2,0; 4,0] ]4,0; 8,0] > 8,0 > 8,0 > 8,0 Intensidade de sujeição a testes sem acidentes prévios There is an

8 Methodology The study design tested whether there is a cause-effect relationship between A&D testing and post-accident reduction, by contrasting the odds of occupational accident risk between workers with different test rates prior to accidents (both exclusively work-related) A&D tests were applied in the workplace at random and by surprise, for 5½ years, after which, it was found whom had accidents and whom had not after n tests (n 0)

test rates prior to accidents (both exclusively work-related) A&D tests were applied in the workplace at

9 Methodology It covered a wide range of data: records concerning accidents, A&D tests or the absence of either one or another 30 biographical and occupational variables for each of (N = 3 801) ever-present employees of a railway transportation company in Portugal, for a period of 5½ years

biographical and occupational variables for each of (N = 3 801)

10 Methodology Homogeneous groups of employees, performing similar tasks and exposed to the same pattern of occupational risks, were studied Within each occupational group, the experimental stimulus of being (or not) tested for A&D constituted a relevant difference The portion untested before any accident, which emerged by chance, became the control group within each occupational risks group

stimulus of being (or not) tested for A&D constituted a relevant difference The portion untested

11 [dentro de Homogeneous groups ] Group 1 (N1 = 3 801): Work onboard trains Group 2 (N2 = 318): Work near or around trains Group 3 (N3 = 1 583): Work away from trains white collars

12 Methodology The methodology applied data-mining techniques (CHAID - Chi-square Automatic Interaction Detector) together with classical statistics hypothesis testing: tests of hypotheses (mean comparisons and analysis of variance) Mann-Whitney Kolmogorov-Smirnov Cramer V Odds ratio all of which with a significance level of 1%

testing: tests of hypotheses (mean comparisons and analysis of variance)

13 Results The CHAID classification trees compared associations between Victim of accident after n tests" and 30 potential explanatory variables, including: Age Sex Academic qualifications Marital status Underage dependents Place of residence Tenure Medical fitness for work Company Business Unit Occupational risk group Shift work rotation Subjection to tests before any accidents Annual test frequency before any accidents

Underage dependents Place of residence Tenure Medical fitness for work Company Business Unit Occupational

15 "Subjection to tests before any accidents" is the most explanatory variable of the dependent variable "Victim of accident after n tests" with very strong association

16 There is a statistically significant difference of victims of accidents among those tested and untested

17 Results Once the expected negative association between accident occurrence and prior tests was confirmed, this study focused on the annual test frequency and accident rates When the initial input variable Subjection to tests was replaced with the time insensitive variable Annual test frequency, the same CHAID algorithm showed that this last variable was the first one next to the top of the tree Again, the testing issue, either expressed only as tested and untested, or expressed in annual frequency, was systematically the most explanatory

variable Annual test frequency, the same CHAID algorithm showed that this last variable was the first one next to the top of the tree

18 [dentro de showed no slide anterior]

19 Results Once the expected negative association between accident occurrence and prior tests was confirmed, this study focused on the annual test frequency and accident rates When the initial input variable Subjection to tests was replaced with the time insensitive variable Annual test frequency, the same CHAID algorithm showed that this last variable was the first one next to the top of the tree Again, the testing issue, either expressed only as tested and untested, or expressed in annual frequency, was systematically the most explanatory

20 Results How far an organisation should go in terms of testing effort? The interest was to find out the optimal frequency, above which there is no benefit in increasing testing, i.e., the frequency of tests at which the accident rates are minimised

21 [dentro de optimal no slide anterior]

22 [dentro da coluna maior (86) no slide anterior] For generality of employees: groups tested for A&D, reported lower accident rates, after any number of tests, than the untested group

24 [dentro da coluna menor (28) no slide anterior] For generality of employees: there is an optimal frequency of testing associated with a minimum accident rate, above which the increase in testing becomes less efficient in terms of prevention

26 Results

27 Results

28 Results

29 Results The results indicate how much more probable is having an accident if untested compared to tested, as being: 3.7 times more, in the sub-population 2.6 times more, in group times more, in group times more, in group 3 The individual's accident risk decreases after being tested

30 Results Optimal testing frequencies that balance testing costs and accident reduction are in the range: ] ] tests per year per worker, in white-collars and professions at large ] ] tests per year per worker, in operations/technical personnel

31 Results The fraction of accident victims that are prevented by the application of optimal frequencies are around: 59% for workers onboard trains 72% for those working near trains 85% for white-collars

32 Results The average costs with application of tests in group of onboard personnel were compared against the money saved from the nonexpenditure with overtime work, due to the reduction of accidents occurred after subjection to tests at the optimal frequency This showed a net saving of about 15 for each 1 invested in testing

33 Conclusions

34 Conclusions Testing for alcohol and drugs at work, has preventive effect in overall professions, stronger in white-collars Each occupational group has an optimal testing frequency associated with a minimum accident rate Testing personnel onboard trains at the optimal frequency generates net savings of at least 15:1 These conclusions emerged from the contrast of accident rates after tests, between homogeneous groups of workers, only differing on their test frequency. Thus, all other things being equal, the different individual frequencies of subjection to testing were likely to be responsible for different outcomes

35 Thank you for your attention Paulo Henriques dos Marques, PhD

AFTRS Health and Safety Risk Management Policy

AFTRS Health and Safety Risk Management Policy Responsible Officer Contact Officer Authorisation Director, Corporate and Student Services Head of Human Resources Chief Executive Officer Effective Date